# Can a butterfly start a hurricane?

Jul 25, 2007 To the editor: "The butterfly effect" is the proposition that the flapping of a butterfly's wings could affect weather half a world away. It was popularized to explain the generation of tropical waves in the Atlantic: a butterfly flapping its wings in equatorial Africa can turn an ordinary tropical wave into the beginnings of a hurricane. Weather forecasting these days is largely about computer modeling, and people like Herb Hilgenberg, the free weather forecaster who goes by the name of his boat, Southbound II, or Bob McDavitt in the South Pacific, who help cruising sailors choose when to make a passage, rely on a favorite model, along with their own observations and those of the sailors they speak to, to make predictions.One scientist who had a major impact on meteorology was Edward Lorenz. In 1959 while on the faculty at MIT's Department of Meteorology, he developed a computer model of an idealized atmosphere to run on the primitive computer in his office. Lorenz plugged a dozen variables into so-called filtered equations, which recalculated them and moved the numbers along in time. It took a minute to simulate one day in the atmosphere, and by plotting the variables Lorenz could draw a line between the points, creating a graph that showed a simple atmosphere changing in a way that repeated itself, but not exactly, much as our weather does. Since a periodic system eventually repeats itself exactly, Lorenz's model was cyclical but non-periodic.

Lorenz wanted to rerun a segment of the model to examine it in greater detail, so he stopped the computer, typed in numbers the computer had calculated several sections back, and started it going. He walked off to get a cup of coffee. By the time he returned the model had generated two months of numbers but the numbers didn't resemble the first run. Lorenz suspected a physical problem with his computer and went back over the machine's trail to locate the mistake. Instead of an obvious break, the first few days were indeed an exact repeat, but after that the model began to generate errors that multiplied until by the end of the second month the graph looked nothing like the original he'd intended to duplicate. At first the numbers differed from the original in tiny amounts, but these kept increasing and doubled in size about every four days.

When Lorenz took the numbers to rerun the model, he'd rounded them off before entering them into the computer. Something in the simple model was amplifying the tiny round-off errors. If the real atmosphere behaved like this, weather prediction would be impossible unless one knew the exact temperature, pressure, humidity, etc. at every point in the atmosphere.

Lorenz knew he was on to something. He called what he discovered in his model, "sensitive dependence on initial conditions." Ultimately, Lorenz came up with the theory of chaos. In simple terms, chaos is the theory of systems that only appear to be random but are in fact predictable if one knows the exact circumstances that generate them. The trouble is that with chaotic systems, even the tiniest difference leads to enormous discrepancies down the line. The atmosphere is chaotic.

Even as meteorologists gain greater understanding of large cycles that drive our weather - El Nino, for example - and the consequences in terms of what we know as weather, the smaller features can bump a system out of its predicted course.

Fortunately for day sailors and racers, short-term forecasts in populated areas are very good. Unfortunately for cruisers, a weather window may only be good for three days. Every day added to the prediction offers more chance for errors to compound.

Ann Hoffner and her husband Tom Bailey voyage aboard their Peterson 44 Oddly Enough.